There is a continuing trend to employ and/or fabricate advanced integrated circuits using techniques, materials and devices that improve performance, reduce leakage current and enhance overall scaling. Semiconductor-on-Insulator (SOI) is a material in which such devices may be fabricated or disposed on or in (hereinafter collectively “on”). Such devices are known as SOI devices and include, for example, partially depleted (PD) devices, fully depleted (FD) devices, multiple gate devices (for example, double or triple gate), and Fin-FET.
One type of dynamic random access memory cell is based on, among other things, the electrically floating body effect of SOI transistors; see, for example, U.S. Pat. No. 6,969,662 (the “'662 patent”). In this regard, the dynamic random access memory cell may consist of a PD or a FD SOI transistor (or transistor formed in bulk material/substrate) having a channel, which is disposed adjacent to the body and separated from the channel by a gate dielectric. The body region of the transistor is electrically floating in view of the insulation layer (or non-conductive region, for example, in a bulk-type material/substrate) disposed beneath the body region. The state of the memory cell is determined by the concentration of charge within the body region of the SOI transistor.
Data is written into or read from a selected memory cell by applying suitable control signals to a selected word line(s), a selected source line(s) and/or a selected bit line(s). In response, charge carriers are accumulated in or emitted and/or ejected from electrically floating body region wherein the data states are defined by the amount of carriers within electrically floating body region. Notably, the entire contents of the '662 patent, including, for example, the features, attributes, architectures, configurations, materials, techniques and advantages described and illustrated therein, are incorporated by reference herein.
Referring to the operations of an N-channel transistor, for example, the memory cell of a DRAM array operates by accumulating in or emitting/ejecting majority carriers (electrons or holes) from body region. In this regard, conventional write techniques may accumulate majority carriers (in this example, “holes”) in body region of memory cells by, for example, impact ionization near source region and/or drain region. In sum, conventional writing programming techniques for memory cells having an N-channel type transistor often provide an excess of majority carriers by impact ionization or by band-to-band tunneling (gate-induced drain leakage (“GIDL”)). The majority carriers may be emitted or ejected from body region by, for example, forward biasing the source/body junction and/or the drain/body junction, such that the majority carrier may be removed via drain side hole removal, source side hole removal, or drain and source hole removal, for example.
Notably, for at least the purposes of this discussion, a logic high data state, or logic “1”, corresponds to, for example, an increased concentration of majority carries in the body region relative to an unprogrammed device and/or a device that is programmed with logic low data state, or logic “0”. In contrast, a logic low data state, or logic “0”, corresponds to, for example, a reduced concentration of majority carriers in the body region relative to a device that is programmed with a logic high data state, or logic “1”. The terms “logic low data state” and “logic 0” may be used interchangeably herein; likewise, the terms “logic high data state” and “logic 1” may be used interchangeably herein.
In one conventional technique, the memory cell is read by applying a small bias to the drain of the transistor as well as a gate bias which is above the threshold voltage of the transistor. In this regard, in the context of memory cells employing N-type transistors, a positive voltage is applied to one or more word lines to enable the reading of the memory cells associated with such word lines. The amount of drain current is determined or affected by the charge stored in the electrically floating body region of the transistor. As such, conventional reading techniques sense the amount of channel current provided/generated in response to the application of a predetermined voltage on the gate of the transistor of the memory cell to determine the state of the memory cell; a floating body memory cell may have two or more different current states corresponding to two or more different logical states (for example, two different current conditions/states corresponding to the two different logical states: “1” and “0”).
Further to writing and reading data to memory cells, data stored in the memory cells is required, under certain circumstances, to be periodically refreshed as a result of leakage current. The refreshing of the memory generally involves periodically reading information or data from an area of the memory (e.g., memory cells), and subsequently rewriting the read information into the same area of memory (e.g., memory cells) from which it was read with no modifications. Conventional refreshing techniques thus use the read and write operations appropriate to the transistor, and perform the read and write during two or more separate clock cycles. The technique used for refreshing data in a dynamic memory can have a large impact on memory performance, including memory availability, die area, and power consumption. Memories are typically and more specifically refreshed by performing a read operation during which data is read from memory cells into sense amps, followed by a write operation during which data is written back into the memory cells.
Conventional solutions to improve memory availability have typically involved increasing the number of sense amps in the memory so more of the memory can be refreshed at the same time. Unfortunately, the addition of more sense amps increases memory die area. Additionally, conventional refresh techniques often lead to relatively large power consumption due to, for example, the separate read and write operations of the refresh. The present inventions, in one aspect, are directed to hold operations that inherently refresh data states of memory cells without the need for separate refresh operations.